(dissenting):
The question I find crucial in this case is whether the district court erred in holding that DOT could not properly conclude on the record before it that the implementation of HM-164 would not “significantly” affect the environment and that an environmental impact statement (EIS) was therefore unnecessary. When we talk of “worst-case” accidents, which one of the two major studies the DOT relied on, the Sandia Report, indicated could result in *7531,800 latent cancer fatalities, 290 early morbidities, and 5 early fatalities, not to mention genetic effects — in the case of plutonium transport — or $9 billion in “Direct Economic Impact” — in the case of PO-0210 transport — it seems almost nonsense to talk of no “significant” impact. “Worst-case” accidents have a way of occurring — from Texas City to the Hyatt Regency at Kansas City, from the Tacoma Bridge to the Greenwich, Connecticut, 1-95 bridge, from the Beverly Hills in Southgate, Kentucky, to the Cocoanut Grove in Boston, Massachusetts, and from the Titanic to the DC-10 at Chicago to the 1-95 toll-booth crash and fire — and that alone would end the case for many. As I said in respect to nuclear accidents in New York City in Morningside Renewal Council, Inc. v. United States Atomic Energy Commission, 482 F.2d 234, 241-42 (2d Cir.1973) (Oakes, dissenting), cert, denied, 417 U.S. 951, 94 S.Ct. 3080, 41 L.Ed.2d 672 (1974), it is the potential environmental effect that is important. In my opinion, the effect of the “worst-case” accident alone would be sufficiently substantial to justify an EIS, since the effect of HM-164 is to permit the transportation of nuclear waste and other materials through the most densely populated city in the United States, when “credible” accidents may occur. But the worst-case accident is only one shade of a spectrum of potentially significant environment impact; as the severity of the accident lessens, and the possibility of its occurrence increases, the potential impact remains serious enough in my view to require an EIS, but I will refer only to this one end of the spectrum, for illustrative purposes.
The DOT concluded that the possibility of an accident involving the release of dangerous radioactive materials is exceedingly low. Specifically, a highest severity (Class VIII) accident, for example, has “a probability nationwide of 1.1 X 10 ~5,” HM-164 at 13, citing Sandia Report, while the probability of such an accident is “estimated to be no greater than the 3 X 10-9 per year for 1975 shipping rates” by the NUREG-0170 Report. Both of these figures — the latter suggesting that a “worst-case” accident could occur only once in 300 million years— are based on 1975 shipments, where presumably there was no worst-case accident. The NUREG-0170 Report attempts to place these figures in perspective by comparing them to “many commonly accepted accident risks” as follows:
Cost in Days of Life Associated With Various Activities (Ref. 3-19)
Cost in
Activity Days of Life
Living in city (rather than in country) .. 1800
Remaining unmarried............... 1800
Smoking 1 pack of cigarettes per day ... 3000
Being 4.5 kg overweight............. 500
Using automobiles.................. 240
170 mrem/year of radiation dose ...... io
Transportation of radioactive material * 0.030
NUREG-0170, at 3-13 (emphasis added).
Individual Risk of Early Fatality by Various Causes (Ref. 5-10)
Number Individual
Accident Type per Year Risk per Year
Motor Vehicle ____5.5x104____ lin 4,000
Falls............1.8x104____ lin 10,000
Fires............7.5x103____ lin 25,000
Drowning........6.2 x 103____ 1 in 30,000
Air Travel .......1.8x 103____ lin 100,000
Falling Objects----1.3 x 103____ l jn 160,000
Electrocution.....1.1 x 103____ i ¡n 160,000
Lightning..........160...... 1 in 2,000,000
Tornadoes .........91....... lin2,500,000
Hurricanes.........93....... 1 in 2,500,000
100 Nuclear
Reactors ......3 x 10-3*.... i jn
5,000,000,000
Transportation of Radioactive Material (from Radioactive 1 in
causes)......3.5x10-4**____ 200,000,000,000***
I find quantifications such as these absurd on their face. The tables above compare the risks of known and fairly common accidents and activities with nuclear related activities for which there is little and limited historical data. Extrapolation on the basis of limited time-place experience is notoriously misleading; I note that the above tables were compiled before the near-cata*754strophic occurrence at Three Mile Island with its sordid tale of human and mechanical error.
I am aware that courts have neither the expertise nor the right to second-guess risk assessment methodologies employed by agencies, see Baltimore Gas & Electric Co. v. Natural Resources Defense Council, Inc., -U.S.-,-, 103 S.Ct. 2246, 2256, 76 L.Ed.2d 437 (1983), and that an EIS need not address remote and highly speculative consequences of proposed action. See, e.g., South Louisiana Environmental Council, Inc. v. Sand, 629 F.2d 1005, 1016 (5th Cir. 1980); Concerned About Trident v. Rumsfeld, 555 F.2d 817, 828-29 (D.C.Cir.1977). But DOT has adopted and is therefore bound by Council on Environmental Quality guidelines enacted for the purpose of implementing NEPA. Those regulations make it clear that, in determining whether a proposed action is likely to affect the environment “significantly,” decisionmakers must consider, among other things, “[t]he degree to which the effects on the quality of the human environment are likely to be highly controversial” and “[t]he degree to which the possible effects on the human environment are highly uncertain or involve unique or unknown risks.” 40 C.F.R. § 1508.-27(b)(4), (5). To be sure, the fact that people are highly agitated and willing to go to court does not alone render a given project “controversial,” see Fund for Animals v. Frizzell, 530 F.2d 982, 988 n. 15 (D.C.Cir. 1975); instead, a project is controversial if “a substantial dispute exists as to the size, nature, or effect of the major federal action.” Foundation for North American Wild Sheep v. United States Department of Agriculture, 681 F.2d 1172, 1182 (9th Cir. 1982) (emphasis in original) (quoting Rucker v. Willis, 484 F.2d 158, 162 (4th Cir.1973). But the purpose of NEPA generally and the EIS requirement in particular is to ensure that the decision-making process canvasses “every significant aspect of the environmental impact of a proposed action.” Vermont Yankee Nuclear Power Corp. v. Natural Resources Defense Council, Inc., 435 U.S. 519, 98 S.Ct. 1197, 55 L.Ed.2d 460 (1978). In this case, as in Baltimore Gas & Electric Co., supra, policymakers must make hard choices “in the face of uncertainties about the long-term effects on the environment.” -U.S. at-, 103 S.Ct. at 2252. Given the “unique [and] unknown risks” associated with and the “highly controversial” nature of nuclear waste transport, I believe that the DOT’s decision not to prepare an EIS was arbitrary, capricious, and not in accordance with law, viz., the CEQ guidelines cited above.
Even if I am wrong with respect to the DOT’s obligation under CEQ guidelines, however, I believe that HM-164 is itself defective because it relies on insufficient or contradictory data at several crucial points and because it fails to address certain risks — the effect of human error and the possibility of sabotage, for example — that bear directly on the possibility of a serious consequence, “worst-case” accident. My conclusion in this regard is based primarily on the following factors which, because they were all addressed by Judge Sofaer in his opinion, I will only summarize here.
1. Shipment volume analysis
The estimates concerning the number of shipments are subject to variations — wholly unresolved by the DOT’s environmental assessment — of several orders of magnitude. The assessment states that
A 1985 shipment model is not contained in the urban studies, but for spent fuel the NUREG 0170 figures are projected from 254 packages shipped by truck per year in 1975 to 1,530 in 1985, a 600% increase. The IRG Report estimates 260 truck shipments of spent fuel in 1985. If the increase used in NUREG 0170 were applied to figures for the number of spent fuel packages and the total number of radioactive materials packages used in the Urban Environs Draft, one would project that 72 truck shipments of spent fuel would occur in New York City in 1985, out of [a] total of 3.95 X 105 packages shipped. (Total packages shipped other than spent fuel are projected in NUREG 0170 for 1985 by applying a factor of 2.6, or a 10% increase per year, to *755the 1975 figures. NUREG 0170, p. A-20; 1978 Urban Environs Draft, p. 61.) The more conservative IRG Report estimate would leave the spent fuel shipments in 1985 in New York City unchanged from 1975.
HM-164 at 19. As Judge Sofaer pointed out,
DOT failed to mention data on spent fuel and plutonium shipments collected in a report prepared for the NRC in 1978, and published by that agency in 1980. That report refers to estimates made by the Environmental Protection Agency (EPA) in a 1974 report on both spent fuel and plutonium shipments with projections through the year 2020. The number of spent fuel shipments (both truck and rail) are estimated to increase to 2,103 by 1985, to 7,513 by 2000, and to 15,043 by 2020. Plutonium shipments (50% by truck) are estimated to increase from 60 in 1975, to 400 by 1985, 4764 by 2000, and 20,737 by 2020. NUREG/CR-0742: Review and Integration of Existing Literature Concerning Potential Social Impacts of Transportation of Radioactive Materials in Urban Areas 71 (Exhibit II-4). The authors wrote:
Although the average number of miles per shipment are presumed to decrease as new facilities are built around the country, the number of shipments per year is estimated to increase so rapidly that total units-times-miles figures increase over the 45 year period by a factor of 18 for spent fuel and 138 for plutonium. Consideration of the number-of-shipments columns might well give one pause, since each shipment provides an opportunity for an accident or a diversion, in addition to possible public response to “normal” transport. The estimated increase from 363 spent fuel shipments in 1975 to over 2,000 by 1985 might be enough cause for concern, but the estimated increase from 60 plutonium shipments to 400 in the same period is potentially much more serious. Id. at 70. Of course, these data might be unreliable, but DOT did not analyze the question.
539 F.Supp. at 1267 n. 9. In my view an agency is no more justified in “negleet[ing] to mention a serious environmental consequence ... or otherwise swe[eping] ‘stubborn problems or serious criticism ... under the rug,’ ” County of Suffolk v. Secretary of Interior, 562 F.2d 1368, 1384-85 (2d Cir.1972), cert, denied, 434 U.S. 1064, 98 S.Ct. 1238, 55 L.Ed.2d 764 (1973) (quoting Silva v. Lynn, 482 F.2d 1282, 1285 (1st Cir. 1973)), in deciding whether to prepare an EIS than it is in preparing an EIS. The failure to resolve the shipment volume question, given its obvious importance in calculating accident risk, strikes me as a serious deficiency in the DOT’s decision-making process.
2. Accidents data
The volume of shipments is but one component of the accident risk estimates relied on in HM-164. The other essential component is accident frequency, i.e., fractional number of accidents per vehicle-kilometer, and I agree with Judge Sofaer’s conclusion that the accident data relied on in HM-164 are wholly inadequate. The NUREG-0170 Report’s truck accident frequency assumption was apparently lifted from a 1974 symposium paper; the Report merely cites this source without discussing either the derivation or reliability of the estimate it contains. The Sandia Report’s estimate of truck accident frequency apparently relies, at least in part, upon the DOT’s own Hazardous Materials Incident Reporting System, a data base that, as Judge Sofaer noted, the DOT itself has conceded is unreliable because
the DOT does not know the exact number of companies subject to this requirement [of reporting accidents;] agency officials concede that those reporting constitute only a small percentage of them. Furthermore, the DOT concedes that the information transmitted by these reports is not reliable. The DOT Hazardous Materials Transportation Task Force noted that “the adequacy and relevancy of *756much of the data [in the incident reports] are questionable” and “the credibility of available incident data is questionable, and there is no routine validation of the data [by the DOT].”
539 F.Supp. at 1268, National Transportation Safety Board, Safety Effectiveness Evaluation 5, citing DOT Task Force Report. HM-164 itself never explicitly addresses this issue, but instead merely observes that “the low historical accident rates do tend to support the research conclusions that the risks in transporting radioactive material by highway are very low.” HM — 164 at 33. That only accidents that are not particularly serious have been reported over the very limited time that the DOT has passively received accident reports submitted from time to time by carriers does not to my mind support the prediction that a serious accident is so unlikely to occur that the preparation of an EIS is unnecessary.
3. Packaging response data
The volume of shipments and the accident frequency rate are the primary components of accident risk; the consequences of an accident, however, turn on the integrity of the package containing the radioactive material. The district court found, and I agree, that HM-164’s treatment of the uncertainty surrounding packaging response data was insufficient. Again, the very sources relied on by the DOT in HM-164 point up the limitations of the data those sources employed. As Judge Sofaer noted, 539 F.Supp. at 1268 n. 11, “NUREG-0170 recognizes [at 5-20] that the ‘paucity of data on package responses to severe accidents’ makes the consequences of such accidents difficult to predict. Accidents became increasingly catastrophic as greater quantities of radioactive materials are released from their containers.” Moreover, HM-164 does not even comment upon the Sandia Report’s statement at 49 that places some of NUREG-0170’s conclusions about severe accidents in doubt:
Before summarizing the results of the analysis, a few comments on data changes since the working draft of this report (Reference 10) was published are in order. The major change has been in the consideration of material which can be released from a spent fuel shipping cask. In the previous analysis, a fraction of the entire cask contents was assumed to be released and aerosolized. After consideration of test data on cask response to severe accidents, the authors decided that this assumption was not reasonable. The analysis of cask accidents in the present study is based upon the more realistic assumption that the major source of radioactive material which would be released in a severe accident is the material deposited on the cask interi- or and exterior surfaces of the fuel elements during their lifespan in the reactor (known as reactor “crud” and containing predominantly primary system corrosion products).... [A]bout 154 curies of Co-60 are present on the surface of the fuel pins. Recent information indicates that approximately 20% of the adhered crud could be easily removed by physical shock.
Given the uncertainties surrounding packaging response — the record indicates that much of the data was derived from computer simulation rather than actual testing— the DOT should have addressed the issue more thoroughly.
4. Human error
In this respect I agree with the district court that human error was not sufficiently factored into DOT’s probability estimates. The majority refers to “the extremely low probability that any cask involved in a truck accident would also be compromised by human error,” majority op. at 35-36, but I note that the Sandia Report bases its analysis on one study referring to 3,939 cask shipments (rail and truck) as to which 16 incidents were “traced directly to a human error or deviation from [quality assurance] practices,” id. at 77, something quite different from the human error entering into transportation accidents themselves. As the report itself concludes:
As mentioned earlier, human errors resulting in accidents are not included in *757this analysis. Although there are possible synergisms that would connect the human error with a vehicular accident, the two were considered separable for this treatment. The results should be interpreted carefully since the source of the initial data for the determination of incident rates were vastly different.
Sandia Report at 84 (emphasis added). Thus I cannot agree with the majority, supra at 36, that “since the effect of human error is inevitably included in the historic rate of accidents, the Department, at least implicitly, considered human error in its probability assessment.” I think it did nothing of the sort, and I do not believe that statistical apologetics in the name of judicial deference to agency expertise is sound as a matter of law or policy.
5. Sabotage
I believe that, under Natural Resources Defense Council, Inc. v. Nuclear Regulatory Commission, 539 F.2d 824, 842 (2d Cir.1976), vacated and remanded on other grounds, 434 U.S. 1030, 98 S.Ct. 759, 54 L.Ed.2d 777 (1978), the DOT’s failure to address the special problems of theft, diversion or sabotage of hazardous nuclear materials is fatal to its conclusion of “no significant environmental effect.” Even though these problems may be “unquantifiable,” Sandia Report at 85, they were thought sufficiently serious to justify fifty-five pages of analysis therein, analysis limited, however, so as “[t]o avoid providing potential adversaries with a ‘cookbook’ of methods.” Id. at 86. The fact that there has been a “world-wide increase in terrorist activity,” id. at 85, that “[h]igh explosives are available commercially in a variety of chemical and physical forms,” id. at 86, and that “[a]ccess to shipments of spent fuel would be possible for an adversary intent upon sabotage or theft,” id. at 91, would seem to make the subject a matter for serious consideration, at least in an Environmental Impact Statement.
I repeat that I fully realize that agencies have to be shown extreme deference in their environmental determinations, particularly when they are “making predictions, within [their] area of special expertise, at the frontiers of science.” Baltimore Gas & Electric Co.,-U.S. at-, 103 S.Ct. at 2256. But for the reasons set out above I believe that DOT’s determination that there is or will be no significant effect upon the environment from a national rule pertaining to the highway routing of radioactive materials was arbitrary and capricious. When we add to that the fact that DOT did not even consider alternative modes of transport or local storage in arriving at its rule, we have a result which effectively negates NEPA and runs contrary to the spirit of cases decided in this court. See, e.g., WATCH (Waterbury Action to Conserve Our Heritage, Inc.) v. Harris, 603 F.2d 310, 326-27 (2d Cir.), cert, denied, 444 U.S. 995, 100 S.Ct. 530, 62 L.Ed.2d 426 (1979); Trinity Episcopal School Corp. v. Romney, 523 F.2d 88, 92-93 (2d Cir.1975); Hanly v. Kleindienst, 471 F.2d 823 (2d Cir.1972), cert, denied, 412 U.S. 908, 93 S.Ct. 2290, 36 L.Ed.2d 974 (1973).
I therefore respectfully dissent.
Calculation based on an average of 0.5 mrem per year to an average exposed individual (see Chapter 4).
Statistical estimate.
Statistical estimate for 1975.
Using a population at risk of 75 million people.